International Population Conference 2017, Cape Town, South Africa - - PDF document

1

International Population Conference 2017, Cape Town, South Africa Session 707 Perspectives on fertility transitions Fertility transition in sub-Saharan Africa: A view using parity progression ratios Mathias Lerch Max Planck Institute for Demographic Research, Rostock, Germany lerch@demogr.mpg.de Thomas Spoorenberg1 United Nations Population Division, New York, USA thomas.spoorenberg@gmail.com Abstract Fertility decline in sub-Saharan Africa was slow and diverged significantly by country. When compared to past experiences in Asia and Latin America, the pattern of decline also seems to be distinct as birth postponement played a major role. We propose a parity-specific perspective of the fertility transitions in 26 countries to identify regularities in the underlying behavioral changes with a focus on the interplay between birth postponement and birth limitation (i.e. the stopping of childbearing). Using multiple World Fertility Surveys and Demographic and Health Surveys, we analyze synthetic parity progression ratios, the singulate mean age at first birth and average birth intervals, as well as the observed and tempo-adjusted levels of total fertility between 1965 and 2010. While early declines in fertility were driven by birth postponement at all parities, we observe a recent onset of family limitation in countries that are most advanced in the fertility transition. Sub-Saharan Africa may thus experience a particular type of transition, in which all parities contribute from the

• utset, and in which the temporal sequencing of stopping and postponement behaviors is

inversed when compared to other world regions.

1 The views expressed in this paper are those of the author and do not necessarily reflect the views of the United Nations.

2

Background and objectives The fertility decline in sub-Saharan Africa (SSA) was late and non-monotonic when compared to other developing regions. The drop in the total fertility rate (TFR) started in the 1970s to 1980s, slowed down in the 1990s and resumed since the 2000s. Although all countries experienced at least some decline, the onset and pace of this trend varied substantially (Garenne 2008). In 2010-2015, SSA still has the highest TFR in worldwide comparison (5 children per woman on average) (United Nations 2017), and there is much uncertainty about the future course of the transitions (Gerland et al. 2017; Schoumaker 2017). We aim at identifying regularities in the national patterns of fertility change over the course

• f the transition in order to help informing future developments. Covering the period between

1965 and 2010 in 26 countries, we analyze fertility according to women’s number of previous births (i.e. parity). This helps to better identify trends in both birth postponement and birth limitation (i.e. the stopping of childbearing). While doing so, we also would like to contribute to the discussion about the exceptionalism of the fertility transition in sub-Saharan African. When compared to past experiences in Asia and Latin America, fertility decline in SSA began at lower levels of economic development and progressed at a slower pace, mirroring the slower change in socioeconomic structures as well as a pro-natalist culture (Bongaarts 2017). Corroborating the predictive conjecture made by Caldwell et al. (1992), the pattern of fertility decline also seems to be distinct. As total fertility dropped, the age-distribution of fertility remained constant. Instead of a classic pattern of family limitation, in which women stop childbearing at progressively lower parities as the fertility transition progresses (Henry 1952), the sub-Saharan Africa pattern has been argued to be mainly driven by birth postponement (Johnson-Hanks 2007; Timaeus and Moultrie 2008; Moultrie et al. 2012). Marriages have been deferred to higher ages from the very start of the fertility transition (Shapiro and Gebreselassie 2014; Hertrich 2017) and all birth intervals have lengthened substantially (Johnson-Hanks 2007; Timaeus and Moultrie 2008; Casterline and Odden 2016). The idea of a target number of children, as well as the intentional stopping of childbearing once this target is achieved, was not part of the pre-transitional cultural repertoire of SSA (van de Walle 1992). The reproductive regime aimed at reproducing the lineage group in a context of high mortality by promoting motherhood through its association with high spiritual and social standing. To protect women’s body from health issues and improve the chances of survival of the youngest child, the reproductive careers have been

3

traditionally characterized by long periods of postpartum insusceptibility (Caldwell and Caldwell 1987; Bledsoe et al. 1998). Birth intervals further lengthened in a context of increasing economic uncertainty (Johnson-Hanks 2007). African fertility transition appears to be essentially determined by postponement of family events. However, recent studies challenge this African exceptionalism. Similar to the experiences in Asia and Latin America, relative fertility decline was more important at higher ages and correlated with the drop in ideal family size (Bongaarts and Casterline 2013). Women are increasingly willing to stop childbearing, especially since 2000 and among higher parities (Casterline and Agyei-Mensah 2017). When compared to experiences in other developing regions, however, fertility preferences are higher in SSA and their decline did not translate in comparable drops in the TFR. The unwanted fertility rate stayed constant over the last 20 years (Günther and Harttgen 2016) because the level of unmet need for contraception is the highest in world-wide comparison. The main difference of sub-Saharan Africa thus lies in the slower diffusion of modern contraceptives, although the trend is upward in all countries (Tsui et al. 2017). Yet in countries with the strongest increase in the prevalence of modern contraceptives (i.e. Southern and Eastern Africa), the purpose of unmet need changed from birth spacing towards the stopping of childbearing (Lesthaeghe 2014). Given this indirect evidence for family limitation in SSA, an analysis of fertility trends by parity is timely. Two previous international inter-cohort comparisons revealed a timid decline in higher order births (after the 5th or 6th birth; Brass et al. 1997; Alter 2016). In the early 1990s, only Zimbabwe, Namibia, Ghana and Kenya had a period parity progression ratio to the sixth birth below 80% after five years of occurrence since the fifth birth (Mboup and Saha 1998). Yet it remains unclear to what extent these slow declines in parity progressions have been driven by birth limitation. Birth postponement also temporarily depresses period measures of fertility (Bongaarts & Feeney 1998). We address this question by documenting the quantum and tempo of fertility decline in SSA. We assess the trend in postponement and investigate whether fertility limitation started to spread across the region and parity groups according to a classic model, from higher to lower parities over time (Henry 1952). Data and method We describe fertility trends between 1965 and 2010 in 26 countries, using data from 113 World Fertility Surveys (WFS) and Demographic and Health Surveys (DHS) (see the list in

4

Appendix; small island states, as well as countries with only one survey, are not considered here). The aim is to identify common parity-specific patterns between countries and by stages

• f national fertility transitions.

Because the decision to have an additional birth depends on the number of children a woman already has (once the fertility transition started), parity measures of fertility development provide useful information to understand better fertility behaviors. Given the importance of the age of the youngest child in the unique fertility transition in sub-Saharan Africa (Moultrie et al. 2012), parity- and duration-in-parity-specific measures are the most indicated to study the national experiences in the region. Compared to the TFR, these indicators are also better suited to detect early changes in childbearing that can announce a new stage in Africa’s fertility development. We use here the synthetic parity progression ratios (SPPRs) which are based on parity- and duration-specific fertility rates (Hinde 1998; Pullum 2004). SPPRs have been applied to describe the changes in fertility and to assess the quality of survey data in several countries,2 but not in sub-Saharan Africa so far. As there is important uncertainty in fertility measurement obtained from different surveys within the same country3 (Schoumaker 2014), we computed survey-specific trends and cross- validated them against one another to increase our understanding of the data problems. We estimated retrospective trends in SPPRs to the first, second, third, fourth, fifth and sixth birth

• ver the 15-year periods preceding each survey. The Nelson-Aalen empirical cumulative

hazard function estimate of the survivor function (by month) was computed for sliding left- and right-truncated 5-year synthetic cohorts in order to smooth out erratic variations. Nulliparous synthetic cohorts were truncated at age 35 and parity-specific cohorts after 10 years have elapsed since the previous birth; we also truncated the cohorts when the at-risk populations fell below 10 women in order to avoid erratic jumps in parity progressions. Based on the survival functions, we estimated the singulate mean age at first birth and the singulate mean birth interval (Hajnal 1953). These indicators of the tempo of childbearing allow to examine the uniqueness of the fertility transition in sub-Saharan Africa.

2 Afghanistan: Spoorenberg (2013a); Albania: Lerch (2013); Central Asia: Spoorenberg (2013b); India:

Spoorenberg (2010); Iran: Hosseini-Chavoshi et al. (2006) and McDonald (2015); Guatemala: Grace and Sweeney (2016); Mongolia: Spoorenberg (2009).

3 Due to biases related to selection of the surveyed population, changing compositions of the samples, as well

as the omission or displacement of births.

5

Cross-validation of the survey estimates Figure 1 illustrates the cross-validation of the survey-specific series of parity progression

• ratios. The progression of nulliparous women to the first birth (in blue), from the second to

the third birth (in red) and from the fifth to the sixth birth (in black) are shown for Ghana─a country that contributes with seven independent survey waves. The estimates from different surveys for overlapping years reveal data problems. The figures for the five years immediately preceding each survey are systematically under-estimated when compared with the retrospective estimates for the same year but derived from subsequent surveys. This can be explained by a well-documented reporting bias (Schoumaker 2014): women and/or interviewers tend to omit recent births or to shift them backward in time in order to avoid filling out the questionnaire module on the health status of recent births. More distant retrospective estimates from each survey, however, align between one another. Figure 1: Synthetic parity progression ratios to 1st, 3rd and 6th births, as estimated from successive surveys, Ghana, 1965-2015. Sources: WFS & DHS. Note: estimates refer to sliding left- and right-truncated 5-year synthetic cohorts which are indexed by the central year. In order to measure the extent to which this quality issue diverges by country and women’s parity, we calculated for each pair of successive surveys the absolute difference between the

6

last estimate derived from the first survey (based on data covering the immediately preceding five years) and the estimate for the same period but based on the second survey. The two green dots in Figure 1 illustrate this point. We then averaged these differences over the survey pairs by country and by parity. The average difference slightly decreases from less than 4 points for the first birth rates to below 3 points for the progression ratio to the second birth (SPPR0->1), and then increases linearly to more than 6 points for the progression ratio to the sixth birth (SPPR5->6; see Figure 6 in the Appendix). The bias is most pronounced in Burkina Faso, Cote d’Ivoire, Ethiopia, Liberia, Lesotho, Madagascar, Namibia and Zimbabwe (at least 6 points difference on average in the SPPR4->5). Absolute differences between two surveys reach up to 16 points for specific parity progressions and countries. This has implication for the training of interviewers. Particular attention should be paid to the reporting and correct dating of higher order births. More fertile women are subject to a higher response fatigue when they complete the birth history module, and interviewers may adopt strategies to limit their work load. However, variations in the composition of the relatively small samples of women at higher parities could also play a role. Given these problems in the estimates, we decided not to consider the information from women’s birth history for the five years immediately preceding each survey. The more distant retrospective estimates are averaged for overlapping years and the trend is smoothed using a Loess function. Table 1 in the Appendix reports the observation periods by country. Estimation of total fertility and tempo effects Based on these smooth estimates of SPPRs, a period measure of total fertility (TF) can be

• btained. We computed the average lifetime parity achieved by the synthetic cohort as a

weighted average of the women’s parities attained, with the weights being the parity- distribution as implied by the chained SPPRs. For the last parity group (women with at least six births), we estimated the attained parity as the average number of children ever born at the survey dates. We then linearly interpolated estimates between surveys (extrapolated for the period before the first survey) and smoothed the trend. As shown in Figure 7 in the Appendix,

• ur estimates of total fertility fit relatively well the United Nations series of TFR (United

Nations 2017), except in the earliest and most recent phases of the fertility transition. When fertility peaked in the 1960s and early 1970s, our survey estimates are systematically below the UN estimates (by 0.5 to 1.5 births). This may derive from the fact that the UN adjusts upward the observed TFRs to ensure coherence with enumerated child cohorts. Most recent

7

UN estimates, by contrast, are situated below our figures which do not include the most recent data from each survey. Thus, our cautious and conservative estimation approach may lead to an under-estimation of the fertility decline. The impact of birth postponement on the fertility trends is assessed by comparing the

• bserved level of total fertility with a tempo-adjusted measure, following Bongaarts and

Feeney’s (1998) approach. The series of TFs were recalculated once the observed SPPRs were adjusted for the change in the singulate mean age at first birth and the singulate mean birth interval. In order to compare the patterns of fertility decline across countries and by stages of the fertility transition, trends are plotted according to the number of years elapsed since the onset

• f fertility decline. Using the UN series of TFRs (annual figures have been linearly

interpolated between the 5-year estimates), the onset is defined as the year when the value peaked last before experiencing a decline of at least 10% (see Table 1 and Figure 7 in Appendix). Results The Figure 2 shows the country-specific trends in the progression ratios to upper parities (from the third to the fourth birth, from the fourth to the fifth, and from the fifth to the sixth). Countries are ranked according to the average annual intensity of the fertility decline (starting with the strongest decline) to facilitate comparison of the behavioural changes in countries that lead and lag in the process of fertility transition. Starting from rank 14 (Uganda), fertility did not yet decline significantly according to our survey estimates. In the large majority of countries, the transition ratios to upper parities remain above 90%. The exceptions are countries in Southern Africa (with ratios below 80%) (i.e. Zimbabwe, Namibia, and Lesotho), as well as Gabon, Ghana, Kenya and Madagascar (slightly below 90%). In these countries, the onset of fertility transition was earlier and our estimation series cover more advanced stages. The 90%- and 80%-thresholds of parity progression have been crossed only in the second and third transition decade, respectively. In countries where the onset of the fertility transition was more recent (i.e. Burundi, Malawi, Zambia, etc.), the progression ratios to the higher parities have remained relatively stable. Yet, if these countries have to follow the same path than those where the fertility transition is

8

more advanced, one can expect significant declines in higher order births in the decades to come. Figure 2: Synthetic parity progression ratios to the 4th, 5th and 6th birth over the course of the national fertility transitions, sub-Saharan Africa, 1965-2010. Sources: WFS & DHS. Note: the estimates refer to 5-year synthetic cohorts indexed by the central year; numbers before the country names indicate the rank in terms of the fertility decline over our

• bservation period; E=Eastern Africa, M=Middle Africa, S=Southern Africa, W=Western

Africa. Despite these timid declines in progression ratios to upper parities, the results do not reveal a classic process of family limitation. There is limited differentiation in the downward drops by

• parity. Only in Zimbabwe is the rate of sixth births significantly lower when compared to that
• f fourth and fifth births; in Lesotho the rates of fifth births are also slightly inferior to the
• ther two progression ratios.

9

Western Africa is the sub-region where very limited fertility changes have been observed. The progression ratios to higher parities have remained high and stable, and seem even to increase in some of the countries of the region (i.e. Chad and Niger). As with fertility at upper parities, the progression ratios to the first, second and third births remain universal (i.e. above 90%) in the majority of countries (Figure 3). The exceptions are again found in Southern Africa, where the ratios fell to levels situated between 90% and 80% (or even lower), as well as in Gabon and Kenya to a lesser extent. Here, we also find more differentiation between the trends at lower parity groups, when compared to upper parities. The declines are steeper in third than in second and first births, which is weak evidence for a pattern of family limitation. Figure 3: Synthetic parity progression ratios to the 1st, 2nd and 3rd birth over the course of the national fertility transitions, sub-Saharan Africa, 1965-2010. Sources: WFS & DHS.

10

Note: the estimates refer to 5-year synthetic cohorts indexed by the central year; numbers before the country names indicate the rank in terms of the fertility decline over our

• bservation period; E=Eastern Africa, M=Middle Africa, S=Southern Africa, W=Western

Africa. As with fertility at upper parities, the trends at lower parities do not indicate any sign for a fertility decline in the countries that are less advanced in the fertility transition. In Liberia, Niger and Chad, we even observe an increase. In order to appreciate the importance of birth postponement, Figure 4 shows the singulate mean ages at first births and the singulate mean birth intervals. In the large majority of countries, the average birth intervals at the onset of the fertility transition were situated between 2.6 and 3 years. Birth intervals lengthened the most (by at least one year) in countries that have experienced a stronger fertility decline. Two striking observations are notable here, as they confirm previous research. In almost all countries the average birth interval does not vary by parity. The pace of increase in birth intervals was also the same across parity groups. (Only in Namibia, did fertility postponement start at higher parities and diffuse to lower parities over time.) However, we observe a recent leveling off in birth postponement in the countries that are more advanced in the fertility transition (e.g. Cameroun, Kenya, Rwanda, Togo, Senegal and Zimbabwe). This recent trend may announce a new stage in the fertility transition of SSA. Yet it may also be related to the stalls of fertility decline observed in some of these countries (even though many observed stalls have been invalidated due to data problems;Schoumaker 2009).

11

Figure 4: Singulate mean age at 1st birth and mean birth intervals over the course of the national fertility transitions, sub-Saharan Africa, 1965-2010. Sources: WFS & DHS. Note: the estimates refer to 5-year synthetic cohorts indexed by the central year; numbers before the country names indicate the rank in terms of the fertility decline over our

• bservation period; E=Eastern Africa, M=Middle Africa, S=Southern Africa, W=Western

Africa. When compared to the timing of higher order births, the mean ages at first birth differed to a greater extent by region at the onset of the transitions. In Western Africa (Senegal, Togo, Ghana, as well as to a lesser extent Nigeria, Benin, Côte d’Ivoire) where nuptiality has traditionally been universal and the earliest (Caldwell et al. 1992), women have continuously postponed their first births. In recent years, women in Western Africa had their first child around the age of 21-22 on average (except in Mali, Burkina Faso and Niger, where the mean ages remains below 20 years). This is three years later when compared to 20-30 years ago. The onset of motherhood was also postponed by about one year in several Eastern African

12

countries, such as Burundi, Kenya and Uganda. Thus Western and Eastern Africa caught up with the Southern region of the continent, which was historically characterized by later marriages (and more extra-marital births). In this latter region, the changes in the timing of first births were limited over the last thirty years. In Figure 5 we compare the observed and tempo-adjusted levels of total fertility to evaluate whether fertility trends have been affected by the postponement of higher order births in Southern and Eastern Africa, and by the later onset of motherhood in Western Africa. Figure 5: Observed and temp-adjusted level of total fertility over the national fertility transitions, sub-Sahara Africa, 1965-2010. Sources: WFS & DHS. Note: the estimates refer to 5-year synthetic cohorts indexed by the central year; numbers before the country names indicate the rank in terms of the fertility decline over our

• bservation period; E=Eastern Africa, M=Middle Africa, S=Southern Africa, W=Western

Africa.

13

During the first transition decade in Kenya, for example, the tempo-adjusted TF remained rather stable but the observed TF declined steadily. This indicates negative tempo effects due to birth postponement. In the two subsequent decades, by contrast, the adjusted TF started to drop alongside the observed decline, indicating that women were indeed reducing the number

• f children they were having (quantum effect). Twenty-five years after the transition’s onset,

the adjusted TF fell even more steeply. As the trend in observed fertility leveled off, the two indicators converged. Thus, the postponement trend slowed down and affected less and less the measurement of fertility. This pattern of an initial tempo-driven fertility decline and a later quantum-driven decline can be observed in all countries that significantly advanced in the fertility transition: in Southern and Eastern Africa, as well as in Ghana and Togo. As the

• nset of the fertility transition is too recent or is yet to be observed in most of the Western

African countries, one cannot draw similar conclusions. Discussion In this paper, we described the parity-specific changes in fertility behaviors in SSA since 1965 and identified regional differences and similarities in order to inform future fertility

• developments. We relied on subsequent surveys to estimate, cross-validate and smooth trends

in synthetic parity progression ratios, and we compared observed and tempo-adjusted levels

• f total fertility. While the analysis confirmed the uniqueness of the fertility transition in sub-

Saharan Africa, it also revealed new patterns that might indicate the start of a more classic fertility decline. In line with previous studies (Moultrie et al. 2012), our results confirm that the initial stage of the fertility transition in SSA has been driven by birth postponement. Given the cultural and socioeconomic diversity within the region, the similarity in the tempo of childbearing between countries is indeed rather impressive. The average birth interval did not vary strongly across the region at the onset of the fertility transitions, and subsequently lengthened at all parities to a similar extent in the large majority of countries. The pace of the postponement of higher order births was stronger in countries which experienced a more pronounced decline in fertility. Inter-regional differences in the average age at first birth have shrunken due to the significant postponement in Western Africa. These trends in birth postponement undoubtedly contributed to improve the health of the mothers and their

14

• children. However, the pace of fertility decline was slow because parity progression ratios

remain very high. This mirrors the persistently low prevalence of modern contraceptive methods in a context of a pro-natalist culture (Caldwell et al., 1992). At the same time, this analysis identified initial signs for a recent onset of birth limitation, especially in the countries that are most advanced in the fertility transition. While the lengthening of birth intervals tended to level off in more advanced stages of the fertility transition (i.e. after thirty years), the quantum of fertility started to decline. Among lower parities, the pace of birth limitation increased with parity, which tends to confirm a classic pattern of family limitation. Yet this association was not found among upper parities. Women with high fertility preferences may be selected in these upper parities, especially in rural

• areas. Compared to the past experiences in Asia and Latin America, the fertility decline in

rural SSA was indeed slower, but the transitions in urban areas progressed at an almost identical pace (Lerch 2017). The results of this analysis point to a particular type of fertility transition in SSA, in which all parities contribute from the outset, and the temporal sequencing of stopping and postponement behaviors is inversed when compared to other world regions. The pattern of fertility decline tended to switch from the postponement to the limitation of births when the TFR reached about five children per women. Future research may analyse this threshold effect more in detail. In early stages of the fertility transition, women may have had more freedom in deciding about the timing of their birth, rather than about family size. As the transition proceeds, the idea of family limitation seems to be increasingly accepted by

• society. This recent onset of birth limitation was observed not only in Southern Africa, but

also in isolated countries in Western and Eastern Africa. One may thus expect stronger fertility decline in the future as the new pattern diffuses throughout these regions. Demographers need to pay close attention to the fertility changes in the coming decades to identify clues that these countries will follow similar paths. References

Alter, George. 2016. Understanding Historical and Contemporary Fertility Transitions: A Birth Interval

• Approach. Paper presented at the IUSSP Seminar on Linking Past to Present - Long-term

perspectives on micro-level demographic processes, 9-11 December 2016, Kashiwa (Japan).

15 Bledsoe, Caroline, Fatoumatta Banja, and Allan G. Hill. 1998. Reproductive Mishaps and Western Contraception: An African Challenge to Fertility Theory, Population and Development Review 24(1): 15-57. Bongaarts, John. 2017. Africa's Unique Fertility Transition, Population and Development Review 43(S1): 39–58. Bongaarts, John, and John Casterline. 2013. Fertility Transition: Is sub-Saharan Africa Different?, Population and Development Review 38: 153-168. Brass, William, Fatima Juarez, and Anne Scott. 1997. An Analysis of Parity-Dependent Fertility Falls in Tropical Africa. In G. W. Jones, R. M. Doublas, J. C. Caldwell & R. M. D'Souza (Eds.), The Continuing Demographic Transition. Oxford: Clarendon Press, pp. 80-93. Caldwell, John C., and Pat Caldwell. 1987. The Cultural Context of High Fertility in sub-Saharan Africa, Population and Development Review 13(3): 409-437. Caldwell, John C., I. O. Orubuloye, and Pat Caldwell. 1992. Fertility Decline in Africa: A New Type of Transition?, Population and Development Review 18(2): 211-242. Casterline, John B., and Samuel Agyei-Mensah. 2017. Fertility Desires and the Course of Fertility Decline in sub-Saharan Africa, Population and Development Review 43(Supplement): 1-27. Casterline, John B., and Colin Odden. 2016. Trends in Inter-Birth Intervals in Developing Countries 1965-2014, Population and Development Review 42(2): 173-184. Garenne, Michel. 2008. Fertility Changes in Sub-Saharan Africa, DHS Comparative Reports 18: 111. Gerland, Patrick, Ann Biddlecom, and Vladimira Kantorova. 2017. Patterns of Fertility Decline and the Impact of the Alternative Scenarios of Future Fertility Change in sub-Saharan Africa, Population and Development Review 43(Supplement): 1-. Grace, Katrhryn, and Stuart H. Sweeney. 2016. Ethnic Dimensions of Guatema’s Stalled Transition: A Parity-Specific Analysis of Ladino and Indigenous Fertility Regimes, Demography 53: 117- 137. Günther, Isabel, and Kenneth Harttgen. 2016. Desired Fertility and Number of Children Born Across Time and Space, Demography 53: 55-83. Hajnal, John. 1953. Age at Marriage and Proportions Marrying, Population Studies 7(2): 111-136. Henry, Louis. 1952. Fécondité des mariages - Nouvelle méthode de mesure, Population 7(4): 697- 700. Hertrich, Véronique. 2017. Trends in Age at Marriage and the Onset of Fertiltiy transition in sub- Saharan Africa, Population and Development Review 43(S1): 112-137. Hinde, Andrew. 1998. Demographic Methods. London: Arnold. Hosseini-Chavoshi, Meimanat, Peter McDonald, and Jalal Mohammad. 2006. The Iranian Fertility Decline, 1981-1999: An Application of the Synthetic Parity Progression Ratio Method, Population 61(5-6). Johnson-Hanks, Jennifer. 2007. Natural Intentions: Fertility Decline in the Africa Demographic and Health Surveys, American Journal of Sociology 112(4): 1008-1043. Lerch, Mathias. 2013. Fertility Decline During Albania's Societal Crisis and its Subsequent Consolidation, European Journal of Population 29(2): 195-220. Lerch, Mathias. 2017. Urban and rural fertility transitions in the developing world: a cohort perspective, MPIDR Working Paper 2017-011. Lesthaeghe, Ron. 2014. The Fertility Transition in Sub-Saharan Africa into the 21st Century, Population Studies Center Research Report 14-838: 23. Mboup, Gora, and Tulshi Saha. 1998. Fertility Levels, Trends and Differentials (Vol. 28). Calverton, Maryland: Macro International Inc. McDonald, Peter, Meimanat Hosseini-Chavoshi, Mohammad Jalal Abbasi-Shavazi, and Arash

• Rashidian. 2015. An assessment of recent Iranian fertility trends using parity progression

ratios, Demographic Research 32(58): 1581−1602 Moultrie, Tom A., Takudzwa S. Sayi, and Ian M. Timaeus. 2012. Birth intervals, postponement, and fertility decline in Africa: A new type of transition, Population Studies 66(3): 241-258.

16 Pullum, Thomas W. 2004. Natality. Measures based on censuses and surveys. In J. S. Siegel & D. A. Swanson (Eds.), The Methods and Materials of Demography (Second Edition ed. San Diego and London: Elsevier Academic Press, pp. 407–428. Schoumaker, Bruno. 2009. Stalls in Fertiltiy Transitions in Sub-Saharan Africa: Real or Spurious?, Université catholique de Louvain, Departement des Sciences de la Population et du Développement, Document de Travail 30: 46. Schoumaker, Bruno. 2014. Quality and Consistency of DHS Fertility Estimates, 1990 to 2012, DHS Methodological Reports 12: 122. Schoumaker, Bruno. 2017. African Fertility Changes. In H. Groth & J. F. May (Eds.), Africa’s Population: In Search of a Demographic Dividend. Cham: Springer International Publishing AG, pp. 197-211. Shapiro, David, and Tesfayi Gebreselassie. 2014. Marriage in Sub-Saharan Africa: Trends, Determinants, and Consequences, Population Research and Policy Review 33(2): 229-255. Spoorenberg, Thomas. 2009. The Impact of the Political and Economic Transition on Fertility and Family Formation in Mongolia, Asian Population Studies 5(2): 127-151. Spoorenberg, Thomas. 2010. Fertility Transition in Indian between 1977 and 2004. Analysis using Parity Progression Ratios, Population-F 65(2): 315-331. Spoorenberg, Thomas. 2013a. An evaluation of the recent fertility changes in Afghanistan: a parity- specific analysis, Journal of Population Research 30(2): 133-149. Spoorenberg, Thomas. 2013b. Fertility changes in Central Asia since 1980, Asian Population Studies 9(1): 50-77. Timaeus, Ian M., and Tom A. Moultrie. 2008. On Postponement and Birth Intervals, Population and Development Review 34(3): 483-510. Tsui, Amy O., Win Brown, and Qingfeng Li. 2017. Contraceptive Practice in sub-Saharan Africa, Population and Development Review 43(Supplement): 166-191. United Nations. 2017. World Population Prospects: The 2017 Revision, Key Findings and Advance Tables: New York, United Nations, Department of Economic and Social Affairs, Population Division. van de Walle, Etienne. 1992. Fertility Transition, Conscious Choice, and Numeracy, Demography 29(4): 487-502.

17

Appendix Table 1: Information on the country-specific data used for this study

Region Country N of Period of observation (central year of 5- year synthetic cohort) & TF Onset of Rank of surveys transition fertility START TFstart END TFend Duration decline Eastern Burundi 2 1974 6.2 2002 6.1 28 1987 19 Africa Ethiopia 3 1987 6.3 2003 5.9 16 1983 9 Kenya 7 1964 7.4 2006 5.4 42 1966 6 Madagascar 4 1979 5.9 2000 5.8 21 1971 18 Malawi 4 1979 6.9 2002 6.5 23 1980 12 Mozambique 3 1984 5.6 2003 5.5 19 1965 17 Rwanda 6 1979 7.3 2006 5.5 27 1978 3 Tanzania 6 1978 6.6 2001 5.6 23 1966 7 Uganda 5 1975 6.8 2003 6.4 28 1985 14 Zambia 5 1979 7.0 2005 6.1 26 1972 8 Zimbabwe 3 1986 5.7 2002 4.3 16 1970 1 Middle Cameroon 4 1978 5.8 2003 5.6 25 1982 15 Africa Chad 3 1983 6.2 2006 7.3 23 1996 26 Gabon 2 1987 5.7 2004 4.6 17 1982 4 Southern Lesotho 3 1991 4.5 2006 3.6 15 1970 5 Africa Namibia 4 1979 5.9 2005 3.6 26 1975 2 Western Benin 5 1968 5.7 2003 6.0 35 1981 21 Africa Burkina Faso 4 1979 6.5 2002 6.7 23 1982 23 Côte d'Ivoire 5 1967 6.5 2003 6.0 36 1972 13 Ghana 7 1966 6.1 2006 5.2 40 1967 10 Liberia 3 1973 5.9 2005 6.4 32 1981 24 Mali 5 1974 6.4 2004 6.6 30 1990 22 Niger 4 1979 6.6 2004 7.5 25 1996 25 Nigeria 4 1977 6.2 2005 6.1 28 1978 20 Senegal 9 1965 6.2 2006 5.9 41 1976 16 Togo 3 1975 6.4 2005 5.8 30 1976 11

Sources: WFS & DHS. Note: the estimates refer to 5-year synthetic cohorts indexed by the central year.

18

Figure 6: Average over countries of the inter-survey differences in the synthetic parity progression ratio estimates, sub-Saharan Africa, 1965-2010. Sources: WFS & DHS. Figure 7: Total fertility according to UN estimates (in red), survey-specific estimates (in blue) and the averaged and smoothed survey trend (in black), sub-Saharan Africa, 1965- 2010. Sources: WFS & DHS. Note: the full and empty red dots indicate, respectively, the years when the UN TFR peaked and when it dropped by 10%. the estimates refer to 5-year synthetic cohorts indexed by the central year.

19